File size: 2,916 Bytes
e5a65c7 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 |
import os
import torch
from torch.utils.data import Dataset, DataLoader
import numpy as np
import cv2
from PIL import Image
import json
import openmesh as om
import pdb
from utils import *
class BiCarDataset(Dataset):
def __init__(self, dataset_folder,input_size=512):
self.dataset_folder = dataset_folder
self.data_index_list = os.listdir(dataset_folder)
self.input_size = input_size
def __getitem__(self, index):
instance_index = self.data_index_list[index]
instance_folder = os.path.join(self.dataset_folder,instance_index)
input_kps= np.zeros(1)
# image/mask/annotation
#processed images and mask
#input_image = cv2.imread(os.path.join(instance_folder,'image','image_reshape512.jpeg'))
#input_mask = cv2.imread(os.path.join(instance_folder,'image','mask512.png'))
#processed image in dataloader
image = Image.open(os.path.join(instance_folder,'image','raw_image.jpeg')).convert('RGB')
polygon,kps,bbox = readjson(os.path.join(instance_folder,'image','annotation.json'))
mask = polygon2seg(image,polygon)
input_image,input_mask,input_kps = reshape_image_and_anno(image,mask,kps,bbox,self.input_size)
# this two function can be used to visualize
#utils.show_seg(nimage,nmask)
#utils.show_kps(nimage,nkps)
#params: shape and pose
beta = np.load(os.path.join(instance_folder,'params','beta.npy'))[:100]
theta = np.load(os.path.join(instance_folder,'params','pose.npy')).reshape(3,24)
#mesh: Here we only read points and uvmap of body only.
#Tbody: T-pose body; Pbody: Posed body.
tmesh = om.read_polymesh(os.path.join(instance_folder,'tpose','m.obj'))
tbody_points = tmesh.points()
tbody_uv = cv2.imread(os.path.join(instance_folder,'tpose','m.BMP'))
pmesh = om.read_polymesh(os.path.join(instance_folder,'pose','m.obj'))
pbody_points = pmesh.points()
pbody_uv = cv2.imread(os.path.join(instance_folder,'pose','m.BMP'))
return {'input_image':input_image,
'input_mask':input_mask,
'input_kps':input_kps,
#'json_annotation':annotation,
'beta':beta,
'theta':theta,
'Tbody_points':tbody_points,
'Tbody_uv':tbody_uv,
'Pbody_points':pbody_points,
'Pbody_uv':pbody_uv
}
def __len__(self):
return len(self.data_index_list)
dataset = BiCarDataset('./3DBiCar')
batch_size = 2
dataset.__getitem__(1)
dataloader = DataLoader(dataset, batch_size=batch_size, shuffle=True)
for batch in dataloader:
for item in batch:
print(item,batch[item].shape)
break
|